Polymorphisms in the Neuropeptide Y Gene and the Risk Of Obesity: Findings from Two Prospective Cohorts

Abstract

Context:

Neuropeptide Y (NPY) increases appetite and food intake in animals. Conflicting evidence has been found for the association between polymorphisms of the NPY gene and obesity in humans.

Objective:

The objective of the investigation was to study four single-nucleotide polymorphisms (SNPs) in the NPY gene [rs17149106 (G>T), rs16147 (C>T), rs16139 (T>C), rs5574 (C>T)] with body adiposity.

Design:

The study design included a candidate gene association study from two cohorts.

Participants:

Two thousand seventy-one women from the Nurses' Health Study and 1268 men from the Health Professionals Follow-Up study participated in the study.

Main Outcome Measures:

Weight and height were self-reported at baseline and updated every 2 yr to calculate body mass index (BMI).

Results:

Two SNPs (rs17149106 (G>T) and rs16139 (T>C)), with minor allele frequencies of 4%, were associated with elevated risks of obesity (BMI ≥ 30 kg/m²) in both cohorts. The pooled adjusted odds ratios [95% confidence interval (CI)] were 1.72 (95% CI 1.20–2.47) and 1.79 (95% CI 1.24–2.60), respectively. Haplotype analyses reflected the associations with the individual SNP. TTCC carriers, with the minor allele of both SNPs, had an increased risk of obesity (odds ratio 1.89; 95% CI 1.29–2.75) compared with those carrying the common haplotype GCTT. Carriers of the rs16139 C allele had greater BMI than noncarriers with a pooled mean difference of +0.58 kg/m² (95% CI 0.01–1.15) among women and men. Both rs17149106 and rs16139 were associated with weight gain since adolescence/early adulthood but were not associated with abdominal adiposity as measured by waist circumference and waist to hip ratio.

Conclusions:

NPY gene variants were significantly associated with weight changes from young adulthood to middle age and with risk of obesity.

Emerging evidence suggests a promising link between neuropeptide Y (NPY) and obesity. The NPY gene, containing four exons, is located on chromosome 7p15.1 and codes for a 36-amino acid peptide that is secreted by neurons (1). Expression of NPY in the arcuate nucleus of the hypothalamus has orexigenic effects. Its injection directly into the central nervous systems of animals leads to obesity through increases in feeding (2). Apart from increases in energy intake, NPY may also play a role in regulating energy expenditure (2, 3). Moreover, a polymorphism of the NPY gene was found to be significantly associated with increased serum total and low-density lipoprotein cholesterol levels (4, 5).

Epidemiological studies on the associations between genetic variability of NPY and obesity-related phenotypes have been mainly focused on one nonsynonymous single-nucleotide polymorphism (SNP) (rs16139), which leads to a single amino acid base substitution (Leu7Pro) in the signal region of prepro-NPY (1). Findings for the association between this SNP and obesity are inconsistent (1, 4, 6–14). Studies found increased body mass index (BMI) among carriers of the rs16139 SNP only among nonobese (1) or non-insulin-resistant (11) individuals and with some heterogeneity in findings between men and women (10, 13).

In the present study, we investigated the association of four variants of the NPY gene (rs17149106, rs16147, rs16139, rs5574) with obesity status and long-term weight change among men and women from two prospective cohorts (Fig. 1). Both rs17149106 (−602 G>T) and rs16147 (−399 T>C) are polymorphisms in the promoter region of the gene, whereas rs16139 (+1128 T>C) and rs5574 (T>C) are changes to the second and third exons, respectively (15). Multiple genetic markers were chosen for this analysis due to recent evidence that these markers may be involved in different aspects of the neuropeptide's expression and provides broad coverage of the gene (15, 16). Understanding the genetic determinants of obesity such as through NPY could possibly lead to novel pharmacological prevention measures or treatments.

Fig. 1.

SNPs typed on the NPY gene. Information on SNP locations are from National Center for Biotechnology Information Building 34 (Bethesda, MD).

Patients and Methods

Study population

In 1976 the Nurses' Health Study (NHS) recruited 121,700 female nurses aged 30–55 yr. Lifestyle, medical, and dietary information was collected every 2 or 4 yr beginning in 1980 (17). Blood samples were collected by mail from 32,826 women in NHS between 1989 and 1990. In 1986 the Health Professionals Follow-Up Study (HPFS) recruited 51,529 U.S. male health professionals aged 40–75 yr. Similar to NHS, ongoing follow-up has been conducted through biennial mailed questionnaires (18). Between 1993 and 1999, 18,159 men in HPFS provided blood samples.

Participants of the present study were nondiabetic controls (2071 women and 1268 men) of two nested type 2 diabetes case-control studies from the NHS and HPFS (19). All participants were Caucasians of European ancestry and provided written informed consent, and the Human Research Committee at the Brigham and Women's Hospital (Boston, MA) approved the study.

Genotyping

Four previously identified NPY SNPs [rs17149106 (C>T), rs16147 (T>C), rs16139 (T>C), rs5574 (T>C)] (15) (Fig. 1) were genotyped in both cohorts. DNA was extracted from buffy coat using a QIAmp blood kit (QIAGEN, Chatsworth, CA). Genotyping was performed using the OpenArray SNP genotyping system (BioTrove, Woburn, MA). Blinded samples typed in duplicate assured the internal quality of the genotyping and resulted in concordance of greater than 99%. The call rates for the SNPs (rs17149106, rs16147, rs16139, rs5574) in the NHS were 97.1, 96.3, 96.4, and 96.9%, respectively, and in the HPFS were 95.7, 95.7, 95.3, and 96.4%, respectively.

Anthropometry assessment

Information on adult weight and height at baseline (1976 for NHS and 1986 for HPFS) and current body weight updated during follow-up were self-reported. To assess the adiposity in early adulthood, the 1980 NHS questionnaire asked about weight at 18 yr of age, and the 1986 HPFS questionnaire asked about weight at 21 yr of age. Weight change since young adulthood was calculated as the difference between weight at age 18 or 21 yr and weight at baseline. For women, the correlation between recalled weight at age 18 yr and documented weight from college or nursing school records was 0.84 (20), and the correlation between self-reported and technician measured adult weight was 0.96 (21). For men, the correlation with recalled and measured weight was 0.97 (21). BMI was calculated based on weight divided by height squared.

In 1986–1987, participants in the NHS and HPFS reported direct measurements of their waists (at the umbilicus) and hips (at the largest circumference) to the nearest quarter of an inch, using a paper tape and detailed measuring directions. The validity of those measurements was also compared against technician measures in men and women (21). For women, the correlations were 0.89 and 0.84 for waist and hip, respectively. For men, they were 0.95 and 0.88, respectively.

Covariate assessment

Information on family history of diabetes (parents and siblings), smoking status, alcohol intake, and physical activity was also collected at baseline for each study and have been previously validated (22). Dietary intake was obtained by semiquantitative food frequency questionnaire (FFQ) in 1980 for the NHS and at baseline (in 1986) for the HPFS. Measures of total energy and macronutrient intake derived from FFQ have been shown to be correlated with repeated dietary records in both men and women (23, 24). The correlations between the FFQ and diet records for total fat, saturated fat, protein, and carbohydrate intake were moderate to strong (0.5–0.7), whereas the correlation was weak for total caloric intake (0.2–0.3) (23, 24). Participants reported on frequency of intake and portion size on food items during the previous year. Nutrient intakes were derived by multiplying the frequency of intake by the average nutrient content of the specified portion size obtained from the Harvard University food composition database, which used U.S. Department of Agriculture data (25) and was supplemented with information from the manufacturer. Energy-adjusted nutrient intakes were used in statistical models according to the residuals method previously described (26).

Statistical analysis

Two measures of linkage disequilibrium, squared correlation coefficient (r²) and Lewontin's standardized disequilibrium coefficient (D′), were computed between pairs of SNPs. Differences in baseline characteristics by genotypes were tested using χ² tests for proportions and simple linear regression for means. Odds ratios (OR) and 95% confidence intervals (CI) on the risk of obesity by genotype were estimated using logistic regression, adjusting for age, physical activity, smoking status, alcohol consumption, coffee intake, and menopausal status combined with hormone replacement therapy (HRT) use among women (premenopausal, postmenopausal without HRT, current HRT, past HRT for women). Estimates from each cohort were then pooled together by meta-analysis averaging the ORs of the individual studies weighted by inverse of variance with a fixed-effects model. No significant between-cohort heterogeneity was indicated by χ² test (P > 0.30). Biennially reported measures of BMI over follow-up (which spanned 1976–2006 for the NHS and 1986–2006 for the HPFS) were used to test for the mean difference in BMI by NPY genotypes and haplotype. Generalized estimating equations (27) with an exchangeable correlation structure were used to account for the correlation between reported measures of BMI.

Haplotype analysis was performed as a way to explore the joint effect of the SNPs on the same chromosome. Haplotype frequencies were estimated using an expectation-maximization algorithm. Estimated haplotype frequencies were used to test for association with obesity risk, adjusting for age and other lifestyle risk factors as described above. Four haplotypes occurring in greater than 1% of the participants in each cohort were identified. We compared individuals with the putative haplotype (TTCC) with that of the most common haplotype in men (GCTT). Among women, this haplotype (GCTT) was only slightly less common (i.e. 0.2% difference in frequency) than the most common haplotype (GTTC), so GCTT was used as reference to maintain consistency between the two cohorts.

SNP-specific associations with anthropometric measures were conducted using linear regression models. For the models that tested for weight/BMI change since young adulthood, models were adjusted for baseline age, lifestyle risk factors, and BMI reported from young adulthood. For the models testing the associations with waist or hip circumference, models were adjusted for baseline age, lifestyle risk factors, and BMI from each cohort at the time of the measurement of waist and hip circumferences. Due to missing data, the sample size was reduced in those analyses involving waist and hip measurements (1479 women, 1081 men). We assessed the significance of associations by a step-down permutation procedure (28), which accounted for multiplicity in testing multiple SNPs as well as multiple phenotypes. The permutation P values were computed based on 10,000 permutation data sets. Genotype and haplotype analyses were conducted using statistical packages SAS (version 9.1 for UNIX; SAS Institute, Cary, NC) and R 2.9.2 (R Foundation for Statistical Computing; www.r-project.org).

Results

Associations of the NPY variants with adiposity in the NHS and HPFS

The minor allele frequencies of the SNPs rs17149106, rs16147 rs16139, and rs5574 (in consecutive order from the 5′ to 3′ end of the gene) were 3.7, 49.0, 3.9, and 48.1%, respectively, in men. The minor allele frequencies were similar to those in women and in HapMap (Utah residents with ancestry from northern and western Europe, CEU) (29). There were no homozygous carriers of the minor allele for rs17149106 in either cohort. Two pairs of SNPs (rs17149106 with rs16139 and rs16147 with rs5574) were in strong linkage disequilibrium (r² = 0.93 and 0.90, respectively; D′ = 0.96 and 0.94, respectively). No significant departures from the Hardy-Weinberg equilibrium in genotype distributions of the four SNPs were detected among controls by χ² test. Carriers of the rs16139 minor C allele did not differ significantly from noncarriers in age, smoking status, family history of diabetes, physical activity, alcohol consumption, or total caloric intake (Table 1). No differences in macronutrient consumption by genotype were found except that women who carried the minor C allele consumed more protein than noncarriers. This association was not observed to replicate among men.

Table 1.

Baseline characteristics by NPY rs16139 genotypes in NHS and HPFS

Variables	Women (NHS)		P value	Men (HPFS)		P value
	TT	CT/CC		TT	CT/CC
n	1914	157		1172	96
Age (yr)	44.1 (7)	43.7 (7)	0.46	55.5 (9)	54.5 (8)	0.30
Current smoking (%)	22.43%	21.15%	0.18	7.27%	8.33%	0.93
Family history of diabetes (%)	20.27%	20.38%	0.97	15.27%	17.17%	0.61
BMI (kg/m2)	23.8 (4)	24.2 (5)	0.20	25.0 (3)	25.7 (3)	0.01
Waist (cm)a	78.9 (11)	81.1 (13)	0.05	94.2 (8)	96.2 (9.6)	0.04
Waist to hip ratioa	0.78 (0.07)	0.78 (0.06)	0.98	0.94 (0.05)	0.95 (0.06)	0.08
Physical activityb	4.1 (3)	4.2 (3)	0.65	21.2 (25)	20.2 (27)	0.69
Coffee (% 4+/d)	23.6%	25.7%	0.89	13.6%	15.8%	0.58
Alcohol intake (g/d)	6.5 (10)	5.4 (7)	0.19	12.1 (15)	12.9 (18)	0.62
Total calories (kcal/d)c	1591 (489)	1570 (460)	0.62	2022 (601)	2121 (655)	0.12
Carbohydrate (% kcal)c	38.6 (9)	37.9 (9)	0.35	47.2 (8)	46.2 (8)	0.25
Protein (% kcal)c	19.1 (4)	20.0 (4)	0.003	18.3 (3)	18.2 (3)	0.81
Total fat (% kcal)c	39.2 (8)	39.4 (8)	0.82	32.0 (6)	32.8 (7)	0.27
Saturated fat (% kcal)c	15.6 (4)	15.9 (4)	0.42	11.0 (3)	11.2 (3)	0.47
Polyunsaturated fat (% kcal)c	5.3 (2)	5.3 (2)	0.91	6.5 (3)	6.5 (3)	0.96
Cereal fiber (g/d)c	2.6 (2)	2.7 (2)	0.47	6.3 (5)	6.3 (4)	0.96

^aAmong participants who provided this information in 1986 for women (n = 1423) and 1987 for men (n = 1087).

^bFor women, physical activity was in hours of at least moderate intensity per week and for men, in metabolic equivalent hours per week.

^cDietary intake reported by FFQ in 1980 for NHS and in 1986 for HPFS.

Minor alleles of both rs17149106 (G>T) and rs16139 (T>C) were associated with an increased prevalence of obesity in men and women. After pooling data from both cohorts, the adjusted OR and 95% CI of obesity for carriers of the rs17149106 variant was 1.72 (1.20–2.47) (P = 0.003), and for the rs16139 variant, it was 1.79 (1.24–2.60) (P = 0.002) (Table 2). The associations remained statistically significant by Bonferonni correction for multiple testing for each SNP and haplotype (i.e. P = 0.05/9 = 0.006). Mean BMI over follow-up was increased among those with the rs16139 minor C allele (pooled mean difference of +0.58 kg/m², P = 0.048) (Table 3). No significant associations by prevalence of obesity or longitudinal BMI were seen with the other two NPY SNPs (i.e. rs16147 and rs5574 SNP). We did not find significant heterogeneity in the associations between men and women (P for tests of heterogeneity >0.20). To ensure that the test for interaction with gender had greater power, individual data were pooled. The P value of the tests for interaction for each SNP was all P > 0.20 as well, suggesting no interactions with gender.

Table 2.

Adjusted OR (95% CI) of obesity risk by NPY SNP and haplotypes in NHS and HPFS in 1986^a

	NHS			HPFS			Pooled
	nb	Adjusted ORa	P value	nb	Adjusted ORa	P value	nb	Adjusted ORa	P value
rs17149106
GG	258/1485	1.00 (reference)	n/a	56/1126	1.00 (reference)	n/a	314/2611	1.00 (reference)	n/a
GT	34/133	1.51 (1.00–2.28)	0.05	10/83	2.64 (1.26–5.52)	0.01	44/216	1.72 (1.20–2.47)	0.003
rs16147
CC	65/402	1.00 (reference)	n/a	16/310	1.00 (reference)	n/a	81/712	1.00 (reference)	n/a
CT	139/787	1.13 (0.81–1.56)	0.47	35/608	1.13 (0.60–2.13)	0.70	174/1395	1.13 (0.85–1.51)	0.41
TT	90/409	1.42 (0.99–2.03)	0.05	15/288	1.08 (0.51–2.27)	0.84	105/697	1.35 (0.98–1.86)	0.07
rs16139
TT	260/1494	1.00 (reference)	n/a	55/1116	1.00 (reference)	n/a	315/2610	1.00 (reference)	n/a
TC/CC	32/115	1.61 (1.05–2.46)	0.03	10/87	2.51 (1.20–5.24)	0.01	42/202	1.79 (1.24–2.60)	0.002
rs5574
CC	98/461	1.00 (reference)	n/a	20/324	1.00 (reference)	n/a	118/785	1.00 (reference)	n/a
CT	142/793	0.83 (0.62–1.10)	0.19	31/614	0.74 (0.40–1.34)	0.32	173/1407	0.81 (0.62–1.05)	0.11
TT	59/353	0.76 (0.53–1.09)	0.13	15/280	0.77 (0.38–1.57)	0.47	74/633	0.76 (0.55–1.05)	0.10
Haplotype	Frequency†
GCTT	45.2%	1.00 (reference)	n/a	47.4%	1.00 (reference)	n/a	47.0%	1.00 (reference)	n/a
GTTC	45.4%	1.15 (0.96–1.39)	0.13	44.5%	0.98 (0.67–1.45)	0.93	45.6%	1.12 (0.95–1.32)	0.19
GCTC	4.0%	0.76 (0.45–1.29)	0.31	3.6%	1.17 (0.45–3.03)	0.75	3.9%	0.84 (0.53–1.34)	0.46
TTCC	3.6%	1.69 (1.10–2.61)	0.02	3.3%	2.65 (1.23–5.72)	0.01	3.5%	1.89 (1.29–2.75)	0.001

n/a, Not available. Bold values are those of statistical significance (P < 0.05).

^aAdjusted for age, smoking, physical activity, coffee, alcohol, and menopausal status for women.

^bCases/controls or haplotype frequency among controls.

Table 3.

Difference in Mean BMI (95% confidence interval) over Follow-Up by NPY genotype in NHS and HPFS

Genotype	NHS		HPFS		Pooled
	Mean difference in BMI (95% CI)a	P value	Mean difference in BMI (95% CI)a	P value	Mean difference in BMI (95% CI)a	P value
rs17149106
GT vs. GG	0.51 (−0.33–1.35)	0.23	0.75 (0.01–1.49)	0.05	0.65 (0.09–1.20)	0.02
rs16147
CT vs. CC	0.09 (−0.47–0.66)	0.75	−0.10 (−0.53–0.34)	0.66	−0.03 (−0.37–0.32)	0.88
TT vs. CC	0.02 (−0.61–0.65)	0.95	0.21 (−0.31–0.73)	0.43	0.13 (−0.27–0.53)	0.52
rs16139
TC/CC vs. TT	0.62 (−0.27–1.50)	0.17	0.55 (−0.20–1.31)	0.15	0.58 (0.01–1.15)	0.048
rs5574
CT vs. CC	−0.03 (−0.56–0.51)	0.92	−0.44 (−0.85 to −0.02)	0.04	−0.28 (−0.61–0.45)	0.09
TT vs. CC	−0.03 (−0.68–0.62)	0.92	−0.38 (−0.86–0.02)	0.12	−0.26 (−0.64–0.13)	0.20

^aMean estimated using generalized estimating equations for repeated BMI measures adjusted for age, time (year), coffee, smoking, alcohol, physical activity, and HRT for women.

In haplotype analysis, the haplotype TTCC [i.e. carrying the minor allele of both rs17149106 (G>T) and rs16139 (T>C)] was significantly associated with increased risk of obesity compared with haplotype GCTT, which reflected the findings of the individual SNP-specific associations. The pooled adjusted OR was 1.89 (95% CI 1.29–2.75) and was significant, even after accounting for multiple testing as described above (P < 0.006). Similarly, using repeated measures of BMI over the follow-up time in each cohort, the haplotype TTCC tended to be associated with a higher mean BMI in women (+0.63 kg/m², P = 0.17) and men (+0.92 kg/m², P = 0.02) compared with common haplotype GCTT (Fig. 2).

Fig. 2.

Mean BMI (±2 se) over follow-up by haplotype TTCC compared with common haplotype GCTT among women (A) and men (B). Generalized estimating equations were used to estimate mean differences in BMI by haplotype, accounting for correlation between measures over time.

Variants of SNP rs17149106 and rs16139 were significantly associated with increased BMI and weight gain since adolescence/young adulthood to baseline compared with the haplotype GCTT in both cohorts (Table 4). On average, carriers of either the rs17149106 or rs16139 gained 4–5 kg more than their counterparts in weight since adolescence/young adulthood. The differences remained significant in pooled analyses after permutation testing to account for testing of multiple phenotypes and genotypes.

Table 4.

Associations for NPY SNP with weight change from young adulthood and abdominal adiposity in NHS and HPFS

	Women			Men			Pooled
	MD (se)a	P value	P′ value	MD (se)a	P value	P′ value	MD (se)a	P value	P′ value
rs17149106
BMI change (kg/m2)b
GT vs. GG	0.51 (0.08)	0.07	0.76	0.73 (0.068)	0.005	0.11	0.58 (0.055)	0.001	0.02
Weight change (kg)b
GT vs. GG	4.80 (0.55)	0.02	0.29	5.36 (0.48)	0.004	0.09	4.98 (0.39)	0.0002	0.0041
Waist circumference (cm)c
GT vs. GG	1.27 (0.28)	0.24	0.99	0.12 (0.26)	0.91	0.99	0.82 (0.20)	0.37	0.99
Waist to hip ratioc
GT vs. GG	0.002 (0.002)	0.76	1.00	0.001 (0.001)	0.82	0.99	0.002 (0.001)	0.71	0.98
rs16147
BMI change (kg/m2)b
CT vs. CC	0.25 (0.09)	0.20	0.97	0.043 (0.08)	0.80	1.00	0.17 (0.06)	0.31	0.99
TT vs. CC	0.18 (0.11)	0.42	1.00	0.28 (0.10)	0.15	0.93	0.21 (0.08)	0.10	0.87
Weight change (kg)b
CT vs. CC	1.42 (0.66)	0.31	1.00	0.26 (0.55)	0.82	1.00	0.97 (0.46)	0.41	0.98
TT vs. CC	1.01 (0.78)	0.52	1.00	1.85 (0.67)	0.18	0.96	1.26 (0.55)	0.15	0.94
Waist circumference (cm)c
CT vs. CC	−0.16 (0.34)	0.83	1.00	0.81 (0.29)	0.20	0.96	0.26 (0.23)	0.40	0.99
TT vs. CC	−0.37 (0.41)	0.65	1.00	0.79 (0.37)	0.29	0.99	0.11 (0.29)	0.63	0.98
Waist to hip ratioc
CT vs. CC	−0.007 (0.002)	0.16	0.95	−0.001 (0.002)	0.75	1.00	−0.004 (0.002)	0.26	0.99
TT vs. CC	−0.009 (0.003)	0.10	0.86	0.002 (0.002)	0.67	1.00	−0.005 (0.002)	0.43	0.97
rs16139
BMI change (kg/m2)b
TC/CC vs. TT	0.47 (0.08)	0.11	0.88	0.58 (0.068)	0.02	0.38	0.52 (0.06)	0.006	0.11
Weight change (kg)b
TC/CC vs. TT	5.83 (0.55)	0.01	0.13	4.32 (0.48)	0.02	0.33	5.28 (0.39)	0.0003	0.01
Waist circumference (cm)c
TC/CC vs. TT	1.17 (0.29)	0.31	0.99	0.33 (0.26)	0.74	1.00	0.82 (0.20)	0.36	0.99
Waist to hip ratioc
TC/CC vs. TT	0.001 (0.002)	0.86	1.00	0.005 (0.002)	0.47	1.00	0.003 (0.001)	0.51	0.97
rs5574
BMI change (kg/m2)b
CT vs. CC	0.18 (0.09)	0.35	1.00	−0.35 (0.079)	0.03	0.47	−0.016 (0.06)	0.32	0.99
TT vs. CC	−0.08 (0.11)	0.73	1.00	−0.22 (0.10)	0.25	0.99	−0.12 (0.08)	0.28	0.99
Weight change (kg)b
CT vs. CC	0.22 (0.64)	0.87	0.98	−2.33 (0.56)	0.04	0.55	−0.71 (0.45)	0.16	0.94
TT vs. CC	−0.83 (0.79)	0.60	1.00	−1.38 (0.68)	0.31	0.99	−1.00 (0.56)	0.27	0.99
Waist circumference (cm)c
CT vs. CC	0.008 (0.33)	0.99	0.99	0.68 (0.31)	0.28	0.99	0.28 (0.23)	0.43	0.98
TT vs. CC	0.38 (0.4088)	0.64	1.00	−0.20 (0.36)	0.79	1.00	0.12 (0.28)	0.90	0.90
Waist to hip ratioc
CT vs. CC	0.0009 (0.002)	0.84	1.00	0.0004 (0.002)	0.97	0.96	0.001 (0.002)	0.87	0.98
TT vs. CC	0.008 (0.003)	0.13	0.91	0.001 (0.002)	0.81	1.00	0.005 (0.002)	0.24	0.99

P′ are permutated P values accounting for multiple phenotypes and genotypes. Bold values are those of statistical significance (P < 0.05).

^aMean difference (se) adjusted for age, smoking, physical activity, coffee, and alcohol (and menopausal status for women).

^bBMI/weight change is difference between baseline BMI/weight (1976 in women, 1986 in men) and adolescence/young adult-recalled BMI/weight (at age 18 yr in women and at age 21 yr in men); Additionally adjusted for BMI at beginning of period (i.e. baseline or in adolescence/young adulthood).

^cSample size was reduced for abdominal adiposity measures, which were reported in1986 for women (n = 1423) and 1987 for men (n = 1087).

Discussion

In two large independent cohorts of middle-aged women and men, the minor alleles of both SNPs rs17149106 (G>T) and rs16139 (C>T) of the NPY gene was consistently associated with increased risk of obesity. Mean BMI among carriers of the rs16139 C genotype was approximately 0.6 kg/m² greater than noncarriers. Carriers of the haplotype TTCC inferred from four SNP (rs17149106. rs16147, rs16139, and rs5574) had almost twice the risk of obesity than those carrying the most common haplotype (GCTT). The NPY SNPs were also associated with greater weight gain since adolescence/young adulthood (age 18 or 21 yr) to cohort baseline when women averaged 44 yr (NHS, 1976) and men averaged 56 yr (HPFS, 1986). However, no consistent differences in dietary intake by genotypes were detected.

The precise molecular mechanisms underlying the observed association between the rs16139 SNP and body adiposity are unclear. Much molecular research has been devoted to NPY after its discovery as an orexigenic neuropeptide (2, 30). Since those initial studies, NPY has also been found to affect energy expenditure independent of food intake (2, 3). Alternatively, it also may have direct effects on adipose tissue, including inhibition of lipolysis as has been demonstrated in human adipocytes incubated with increasing concentrations of NPY in a dose-respondent manner (31).

The rs16139 SNP is a variant that leads to a change from leucine to proline in the amino acid sequence of its propeptide (1). This change in the signal sequence has been attributed to altered packaging of the hormone in granules of endocrine cells, resulting in higher levels of peptide secretion (32) without affecting the peptide's binding to its receptors (1). Heterozygous carriers of this functional mutation have lower basal concentrations of NPY at rest (33) but greater plasma NPY during sympathetic stimulation such as from exercise (34). The difference in findings suggests that increased storage of the peptide at rest leads to its exaggerated release under stimulation (33). Although the mechanism for increased BMI associated with the NPY variant is thought to be through greater food consumption among carriers than noncarriers, little evidence from epidemiological studies has been found to support this hypothesis. As with our studies, others (4, 6, 35) have not found differences in dietary intake associated with the rs16139 SNP despite findings from animal models (2). The genetic effects on diet consumption may be moderate and not detectable due to misclassification of diet by use of dietary recall in humans. Moreover, if the increase in NPY release occurs only under sympathetic stimulation as previously suggested (33), more episodic exaggerated intakes (e.g. binge eating) rather than a constant increase in food consumption may be more relevant. Further studies are needed to understand the molecular mechanisms underlying the observed associations.

Previous studies found inconsistent associations between rs16139 SNP and obesity (1, 4, 6, 8–10, 12–14). Findings from the present study are in line with those suggesting a significant association between this SNP and adiposity, especially among nonobese individuals. For instance, carriers of the rs16139 C allele had an increased BMI by about 1 kg/m² among Swedish men (n = 510) and women (n = 396) who were not obese. However, the difference in BMI was not observed among individuals who were obese, and the authors suggested that other factors after the onset of obesity may override the effects of the NPY mutation among obese participants (1). Evidence that NPY may not cause extreme obesity has been shown recently. In a large association study among 1842 European-American and 1031 Polish participants, no replicable associations between 26 SNP from the NPY gene and risk of obesity defined by BMI between the 90th and 97th percentile of the distribution compared with lean individuals defined by BMI between the fifth and 12th percentiles (8). The OR for the risk of obesity were 0.83 (95% CI 0.56–1.25) among men and 0.97 (95% CI 0.68–1.38) among women for the rs16139 mutation (8). Biologically, animal studies have found that leptin strongly controls NPY expression (36). Therefore, among obese participants who are leptin resistant, it is plausible that the genetic regulation of NPY expression is overcome by the influences of leptin (36). The majority of participants in both the NHS and HPFS are nonobese (72% women, 95% men) and hence less likely to be leptin resistant.

Although the rs16139 SNP has been the most well studied in human populations, there are other NPY polymorphisms that may also play a role, such as rs16147, which is also a functional variant regulating transcription (37). Carriers of the T allele for rs16147 have been found to have significantly higher circulating levels of NPY (46.3 vs. 41.0 pmol/liter) among participants who previously underwent cardiac catheterization (16). Although we did not find rs16147 to be associated with increased BMI or obesity status in our cohorts, some evidence from East Asian populations [who do not carry the L7P mutation (38, 39)] has supported its role in ischemic stroke (40).

The finding that carrying the NPY SNP variants are associated with greater gain in weight and BMI from young adulthood in both cohorts suggests that early adulthood through middle age is a critical period for the expression of genetic effects. Of note, though, is that the association, although significant, was rather small with a combined weight gain of approximately 5 kg over 20–30 yr. The genotype distribution of rs16139 did not differ among 272 Dutch men (average age 28 yr) who were high weight gainers (11 kg over 7 yr) compared with those who were weight stable (±0.5 kg) (13). Extreme weight gain, therefore, may be under influences of other factors.

There were some limitations to our study. The low frequency of the risk alleles led to limited power, which may have affected the precision of our risk estimates. There may have also been limited ability to test for small genetic effects on dietary intake due to imprecise recall. Reliance on self-reported weight may have also allowed for some measurement error. However, the responses have been previously validated in both studies and have strong correlations with technician measured values (20, 21). The cohorts were also composed on mostly nonobese men and women who are not representative of the obesity level of the current U.S. population. Our participants had higher education and were particularly healthier in general. This may have affected the generalizability of our results to the broader U.S. population. It should be noted that even among this relatively healthier and lean population, the NPY variants were significantly associated with body adiposity and weight gain. We also did not have data on how the haplotype TTCC may have affected the expression of NPY. Lastly, measurement error on total calories may preclude any definitive conclusions of the genetic effects on total energy consumption and expenditure. As for strengths, our study is among one of the largest to date specifically looking at NPY and obesity. We had longitudinal measures of BMI and information on important risk factors. Moreover, the present study was conducted in two independent cohorts. The associations of the NPY genotypes and haplotypes with adiposity were highly consistent in the study populations, suggesting that our findings are less likely due purely to chance. The homogeneity of our population, with only Caucasians of European descent, may have also helped with detecting a genetic association by decreasing the likelihood of population stratification in biasing the observed associations (41).

In conclusion, the NPY variants rs16139 and rs17149106 are associated with increased BMI and risk of obesity among both women and men of Caucasian European descent. The functional rs16139 mutation is also related to weight gain from early age. Further studies are needed to elucidate the molecular mechanisms underlying the observed associations.